International Journal of Refrigeration 24 (2001) 73±87

www.elsevier.com/locate/ijrefrig

Experimental investigation on condensation heat transfer

and pressure drop of new HFC refrigerants
(R134a, R125, R32, R410A, R236ea)
in a horizontal smooth tube
A. Cavallini a,*, G. Censi a, D. Del Col a, L. Doretti a,
G.A. Longo b, L. Rossetto a
a
Dipartimento di Fisica Tecnica, UniversitaÁ di Padova, via Venezia 1, I-35131 Padova PD, Italy
b
Dipartimento di Tecnica e Gestione dei Sistemi Industriali, UniversitaÁ di Padova PD, Italy

Dedicated to Professor Dr.-Ing. Dr. h.c. mult. Karl Stephan on the occasion of his 70th birthday

Received 22 May 2000; received in revised form 2 August 2000; accepted 2 August 2000

Abstract
This paper reports experimental heat transfer coecients and pressure drops measured during condensation inside a
smooth tube when operating with pure HFC refrigerants (R134a, R125, R236ea, R32) and the nearly azeotropic HFC
refrigerant blend R410A. Data taken when condensing HCFC-22 are also reported for reference. The experimental
runs are carried out at a saturation temperature ranging between 30 and 50 C, and mass velocities varying from 100 to
750 kg/(m2 s), over the vapour quality range 0.15±0.85. The e€ects of vapour quality, mass velocity, saturation tem-
perature and temperature di€erence between saturation and tube wall on the heat transfer coecient are investigated
by analysing the experimental data. A predictive study of the condensation ¯ow patterns occurring during the tests is
also presented. Finally comparisons with predictions from the model by Kosky and Staub (Kosky PG, Staub FW.
Local condensing heat transfer coecients in the annular ¯ow regime. AIChE J 1971;17:1037) are reported for all the
data sets. # 2000 Elsevier Science Ltd and IIR. All rights reserved.
Keywords: Heat transfer; Condensation; Pressure drop; Refrigerant; R134a; R125; R410A; R236ea; Smooth tube; Horizontal tube;
Heat transfer coecient; Measurement

Etude expeÂrimentale sur le transfert de chaleur lors de la

condensation et sur la chute de pression des nouveaux
frigorigeÁnes HFC (R134a, R125, R32, R410A, et R236ea)
dans un tube lisse horizontal
ReÂsumeÂ
Cette communication deÂcrit les coecients de transfert de chaleur et les chutes de pression mesureÂes lors de la con-
densation aÁ l'inteÂrieur d'un tube lisse lorsqu'on utilise des frigorigeÁnes HFC purs (le R134a, le R125, le R236ea et le R32)

* Corresponding author. Tel.: +39-049-827-6890; fax: +39-049-827-6896.

E-mail address: alcav@dft.unipd.it (A. Cavallini).

0140-7007/01/$20.00 # 2000 Elsevier Science Ltd and IIR. All rights reserved.
PII: S0140-7007(00)00070-0
74 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87

et le R410A, un meÂlange HFC quasi azeÂotropique. On fournit des donneÂes obtenues lors de la condensation de HCFC-22
a®n de servir de reÂfeÂrence. Les essais ont eÂte e€ectueÂs aÁ une tempeÂrature de saturation allant de 30 aÁ 50 C, pour des deÂbits
massiques dans un eÂventail de 100 aÁ 750 kg/(m2.s) pour une gamme de qualite de vapeur de 0,15 aÁ 0,85. A l'aide de
l'analyse des donneÂes expeÂrimentales, les auteurs ont eÂtudie les e€ets de la qualite de la vapeur, le deÂbit massique, la tem-
peÂrature de saturation et la di€eÂrence de tempeÂrature entre la tempeÂrature de saturation et la tempeÂrature de la paroi sur le
coecient de transfert de chaleur. On preÂsente eÂgalement une eÂtude preÂvisonnelle de l'eÂcoulement lors de la condensation.
En®n, les auteurs e€ectuent des comparaisons avec les preÂvisions du modeÁle de Kosky et Staub (Kosky PG, Staub FW.
Local condensing heat transfer coecients in the annular ¯ow regime. AIChE 1971;1037) pour toutes les donneÂes
obtenues. # 2000 Elsevier Science Ltd and IIR. All rights reserved.
Mots cleÂs : Transfert de chaleur ; Condensation ; Perte de pression ; FrigorigeÁne ; R134a ; R125 ; R410A ; R236ea ; Tube lisse ; Tube
horizontal ; Coecient de transfert de chaleur ; Mesure

Nomenclature  heat transfer coecient (W mÿ2 Kÿ1)

" void fraction
A inside surface area (m2)  density (kg mÿ3)
c speci®c heat (J kgÿ1 Kÿ1)  shear stress (N mÿ2)
d inside diameter (m)  viscosity (kg mÿ1 sÿ1)
G mass velocity (kg mÿ2 sÿ1)
h enthalpy (J kgÿ1)
hLG latent heat (J kgÿ1) Subscripts
JG dimensionless vapour mass velocity a acceleration
Nu Nusselt number av average
p pressure (Pa) CALC calculated
Pr Prandtl number EXP experimental
Re Reynolds number f friction
T temperature (K) in inlet
u velocity (m sÿ1) G gas phase
ut friction velocity [ut=(t/L]0.5 L liquid phase
x vapour quality m mean
Xtt Martinelli parameter PREC precondenser
Xtt ˆ …L =G †0:1 …G =L †0:5 ‰…1 ÿ x†=xŠ0:9 sat saturation
z axial coordinate (m) sup superheated

T temperature di€erence between saturation w tube wall
and tube wall (K)

x vapour quality change in the test section

1. Introduction two-phase regions of the equipment. For those reasons,

Air cooled and water-cooled condensers with in-tube
condensation of refrigerants are widely used in air-condi-
tioners and heat pumps. The substitution of traditional
refrigerants, and particularly of R22, calls for the use of
new refrigerants or zeotropic mixtures of HFCs with
zero ODP (ozone depleting potential). The new pure,
azeotropic and zeotropic refrigerants are now under
scrutiny to show their heat transfer and pressure drop
performance in smooth and enhanced tubes. When
considering zeotropic mixtures, a heat transfer degra-
dation during condensation builds up as compared to
pure ¯uids or azeotropic refrigerants, due to mass
transfer di€usion, sensible heating e€ects and non-equi-
librium due to the separation of the phases. Besides, a
composition shift occurs from leakages taking place in
the recourse to pure or azeotropic refrigerants is prefer-
able in most applications.
The goal of the research work in this ®eld is to max-
imize heat transfer while minimizing pressure drop.
When the substitution for traditional ozone-depleting
refrigerants involves the use of higher pressure ¯uids,
this usually implies a reduction of friction losses, with
higher energetic performance, and a reduction of the size of
the equipment. Thus, heat transfer coecient on its own
does not give enough information to characterize the heat
transfer performance of new refrigerants. Accurate heat
transfer and pressure drop data are also requested to vali-
date existing heat transfer models and/or to develop new
predictive correlations or analyses of the phenomenon.
The aim of this work is to present a complete set of
data relative to condensation heat transfer and pressure
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
drop inside a smooth horizontal tube for new HFC
refrigerants and for R22, whose information provides a
reference for discussion on its HFC candidate sub-
stitutes. These tests cover a wide range of mass velocities
and operating pressures, as they regard low pressure
(R236ea), mid-pressure (R134a, R22) and high pressure
(R32, R125, R410A) ¯uids. A few semi-empirical corre-
lations are available in the open literature for the pre-
diction of condensation heat transfer coecient, but no
method usually displays a general applicability. The
present authors show the predictability of the shear-
dominated heat transfer coecients for such new ¯uids
by relating the heat transfer to the friction pressure
gradient, as shown in the past by Kosky and Staub [1]
and by Traviss et al. [2]. This method holds true only
when a condensate annular ¯ow pattern is present. It is
well known that di€erent types of analyses are necessary
for prediction in di€erent condensation ¯ow regimes
and the failure to select the proper ¯ow pattern can
result in severe errors. Thus a di€erent procedure has to
be applied for gravity dominated condensation heat
transfer data.
2. Literature survey
New refrigerants have been tested during condensa-
tion inside a smooth horizontal tube by several authors.
Nowadays most reported heat transfer coecients are
quasi-local, obtained when small quality variations
occur in the test-tube. This experimental procedure
helps with the understanding of the physical phenomenon.
Therefore, only the most recent works that present local
heat transfer data will be reviewed here.
Dobson and Chato [3] report data taken with refrig-
erants R22, R134a, R410A, R32/R125 (60/40% by
mass) inside smooth round horizontal tubes with inner
diameters ranging from 3.14 to 7.04 mm. The authors
observed ¯ow regimes and developed criteria for transi-
tions. They also presented two correlations able to pre-
dict the heat transfer coecient in the stratifying regime
and in the annular one respectively. Unfortunately some
¯uid thermophysical properties, used by the authors
when developing their own models, are now obsolete
and new quite di€erent values are suggested by a recent
database (Refprop 6.01 [4]).
Tang [5] reports quasi-local data taken with R22,
R134a, R410A, condensing inside a smooth tube with
8.8 mm inner diameter. Tang [5] also developed a new
heat transfer correlation valid in the annular ¯ow
regime, with mass velocities greater than 300 kg/(m2 s)
and reduced pressures between 0.2 and 0.53.
Ebisu and Torikoshi [6] report measurements with
R22, R407C, R410A in a 6.4 mm inner diameter tube,
showing satisfactory agreement with predictions from
the Haraguchi et al. [7] correlation. Eckels and Tesene
75
[8] report data for R22, R134a, R410A, R407C ¯owing
in a 8 mm inner diameter tube and compare experi-
mental data with predictions from existing correlations.
During condensation inside horizontal tubes, the two-
phase ¯ow pattern may be dominated by vapour shear
or gravity forces. While annular ¯ow pattern is asso-
ciated with high vapour shear, stratifying, wavy and
slug ¯ows appear when gravity is the controlling force.
In a fully developed annular ¯ow pattern, there is a thin
uniform condensate ®lm on the entire tube wall, while
the gas phase ¯ows in the central core, and heat transfer
is governed by vapour shear and turbulence. In the open
literature, several theoretical condensation heat transfer
models in annular ¯ow processes have been developed,
among which the semi-empirical equations by: Cavallini
and Zecchin [9], Shah [10], Tang [5] and the theoretical
analyses by Traviss et al. [2] and Kosky and Staub [1].
In stratifying ¯ow processes, a thick condensate layer
¯ows at the bottom of the tube and a thin liquid ®lm
settles on the wall in the upper portion of the tube. Heat
transfer through the thin ®lm may be analysed by the
classical Nusselt [11] theory, as suggested by Jaster and
Kosky [12].
Two additional computational procedures, the model
by Dobson and Chato [3] and the model by Haraguchi
et al. [7] cover both shear and gravity controlled con-
densation phenomena, inside smooth tubes.
Some of these models are compared in Cavallini et al.
[13] against data by present authors and from other
sources. The comparison shows that quite a few of the
available experimental points fall outside of the validity
ranges given for the predicting procedures. This mainly
happens with the so called new high pressure HFC
refrigerants, such as R125, R410A and the quasi-azeo-
tropic mixture R32/R125 (60/40% by mass). Practically
no experimental points fell within the given validity
ranges of the Cavallini and Zecchin equation and of the
Haraguchi et al. model. It can also be seen that predic-
tion by the Dobson and Chato model is unsatisfactory
for the high pressure ¯uids at high values of the Nusselt
number. Better predictions are obtained by making use
of the Shah model, and even better predictions are supplied
by the Tang equation; nevertheless both those models dis-
play limited application within their set validity ranges.
Regarding the data for refrigerants which are usually
applied at a lower reduced pressure, called here low and
mid-pressure refrigerants (R236ea, R134a, R22), the
predicting calculations reported by Cavallini et al. show
generally a satisfactory agreement, although some of the
models display somewhat too limited validity ranges with
respect to today's application requirements. Cavallini et
al. [13] conclude that future work is certainly needed so
as: (a) to extend the validity of existing correlations and/
or to develop new predicting models; (b) to better vali-
date existing theoretical analyses or/and to develop a
new analysis of the phenomenon.
76 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
Better results are found by Cavallini et al. [14] when
predicting the condensation heat transfer coecients for
high pressure ¯uids by using the Kosky and Staub ana-
lysis, along with the Friedel correlation for two-phase
pressure drop.
3. Test facility
The present experimental investigation is carried out
on a new apparatus built at the Dipartimento di Fisica
Tecnica, UniversitaÁ di Padova, for heat transfer mea-
surements during condensation of pure refrigerants and
refrigerant mixtures inside smooth and enhanced tubes.
It allows in-tube condensation heat transfer and pres-
sure drop data to be collected over a wide range of
operating conditions. The facility has a maximum
working pressure of 4 MPa, while refrigerant ¯ow rates
can be varied up to 750 kg/(m2 s).
3.1. Test rig
The purpose of the test rig is to supply a working ¯uid
to the inlet of the test section at a variety of thermo-
dynamic states. A schematic of the apparatus is pro-
vided in Fig. 1.
The experimental facility consists of three loops: the
refrigerant loop, the cooling water loop and the hot
water loop. In the ®rst loop the refrigerant passes
through the boiler and the superheater. Superheated
vapor comes out and then partially condenses in the
precondenser to achieve the set quality at the inlet of the
test section. The refrigerant enters the tube side of the
Fig. 1. Schematic view of the refrigerant loop.
Fig. 1. ScheÂma du circuit frigori®que.
test section at a known mass velocity and vapor quality
and then it is condensed against the cold water ¯owing
in the annulus. The two-phase mixture leaves the test
section and goes to an after-condenser, a braised plate
type condenser, where it is fully condensed and sub-
cooled. The subcooled liquid is passed through a drier
and then sent back to the boiler by a variable-stroke
volumetric pump. A bladder accumulator connected to
a nytrogen bottle and a pressure regulator are installed
in the refrigerant loop.
The refrigerant ¯ow can be independently controlled
by the volumetric pump. The inlet vapor quality to the
test section is determined by the heat extracted in the
precondenser, which can be controlled varying the water
temperature and ¯ow rate. At the same way, the heat
transfer rate in the test section can be varied by adjust-
ing the temperature and ¯ow rate of the cooling water.
Thus vapor quality, mass velocity, and heat ¯ux can be
independently controlled.
The boiler and superheater are both coaxial tube-in-
tube heat exchangers, in which refrigerant ¯owing on
the tube side is vaporized and then heated by hot water
¯owing in the annulus side. In the hot water circuit, an
electrical heater supplies the water with the power
transferred to the refrigerant. A by-pass system is used
to vary the water ¯ow rate. Water ¯ow rates in the two
heating sections are measured by magnetic type ¯ow
meters.
The precondenser and the test section are also tube-
in-tube heat exchangers. The temperature of the cooling
water entering the precondenser and the test section is
controlled by two secondary water circuits. An electrical
heater provides hot water while cold water is provided
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
by a chiller. The cold and hot circuit ¯ows are mixed
together in a storage tank, where the water temperature
is maintained constant. Thermostatic control of the
water is accomplished by adjusting the electrical power
supplied to the heater.
3.2. Test section
The test section, which is a counter ¯ow tube-in-tube
condenser, is shown in Fig. 2. The refrigerant condenses
in the inner tube while the cooling water ¯ows in the
annulus. The test section is divided into two di€erent
parts: a pre-conditioning section around 300 mm long,
where the refrigerant ¯ow, under condensation, achieves
a fully developed ¯ow regime, and the measuring sec-
tion, around 1.0 m long, in which the heat transfer
coecient is measured. This structure is obtained using
a single inner tube 1.6 m long and two separate cooling
water jackets, connected in series on the water side.
The inner tube, which is the test tube, is instrumented
with eight copper-constantan (type T) thermocouples
embedded in its wall to measure surface temperature.
The thermocouples are inserted and soldered into four
equidistant axial groves. Four thermocouples are loca-
ted 100 mm past the inlet of the measuring section. The
other four thermocouples are located 100 mm before the
outlet. Angular locations for the eight thermocouples
are depicted in Fig. 2.
Refrigerant temperatures at inlet and outlet of the test
section are measured by means of adiabatic sections,
Fig. 2. Drawing of the test section [length in (mm)].
Fig. 2. ScheÂma de la section d'essai [longueur en (mm)].
77
with thermocouples inserted into both the refrigerant
¯ow and the tube wall.
The water temperature gain in the test tube is mea-
sured by a four junction copper-constantan thermopile
inserted into appropriate mixing chambers. A metal helix
is wound around the test tube inside the annulus, in order
to avoid strati®cation in the water ¯ow and to get high
heat transfer coecient on the water side together with
appropriate values for the water temperature rise.
Two digital strain gauge pressure transducers (abso-
lute and di€erential transducers) are connected to
manometric taps to measure the vapour pressure
upstream and downstream of the test tube. The pressure
losses are measured in the 8 mm inside diameter, 1.6 m
long, smooth tube by a di€erential transducer.
4. Data reduction and experimental uncertainty
The heat ¯ow rate exchanged in the test tube is
derived from an enthalpy balance on the cooling water
side. The average condensation heat transfer coecient
is obtained as:
 ˆ q=…A
Tln † …1†
where q is the heat ¯ow rate exchanged in the tube, A is
the exchange surface area and
Tln is the logarithmic
mean temperature di€erence between the vapour and
the tube wall.
78 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
The vapour quality entering the test section (xin) is
calculated from an energy balance on the precondenser.
For a pure refrigerant or an azeotropic mixture, the
vapour quality change is given as a ratio of the isobaric
change in enthalpy in the test section
h to the latent
heat hLG. The following two equations are used to cal-
culate the inlet vapour quality and its average value in
the test section xm:
ÿ


hPREC ˆ hsup ÿ hdew ‡ hLG …1 ÿ xin †
…2†
xm ˆ xin ÿ
h=…2 hLG †
where
hPREC is the speci®c enthalpy variation in the
precondenser, hsup is the speci®c enthalpy of super-
heated vapour at the inlet of precondenser and hdew is
the speci®c enthalpy of saturated vapour at the local
pressure. Change in enthalpy in the test section (
h) is
obtained from the heat ¯ow rate and the refrigerant
mass ¯ow rate, which is measured by a Coriolis e€ect
mass ¯ow meter having an accuracy of 0.4% of the
measured value and inserted downstream of the pump.
Experimental heat transfer coecients are reduced
from the measured values of saturation temperature and
wall temperature. All the thermocouples were calibrated
using a bath with a Kaye reference thermometer (accu-
racy 0.02 C). The average accuracy for the thermo-
couples is estimated to be equal to 0.05 C. For the
thermopiles, the average accuracy is around 0.03 C.
Comparison between ¯ow temperature and wall tem-
perature inside the adiabatic sections gives a maximum
deviation of 0.04 K.
The refrigerant pressure is measured at the inlet of the
test section by a digital strain gauge transducer with an
accuracy of 2 kPa for low pressure ¯uids and 7 kPa
for high pressure ¯uids. This measured value along with
the condensation curve by Refprop [4] provides the
saturation temperature, which is compared to the tem-
perature directly measured. For pure ¯uids, the two
values di€er by less than 0.1 C at high mass velocity [G
5 400 kg/(m2 s)]. Higher deviations are found at low
mass velocity [G=100 kg/(m2 s)] and low vapour qual-
ity, and are probably due to non-equilibrium in the ¯ow
and to local sub-cooling at the outlet of the test section.
Table 1
Accuracy for sensors and parameters
Tableau 1
PreÂcision des capteurs et parameÁtres
Temperature  0.05 C
Refrigerant ¯ow rate  0.4%
Water ¯ow rate  1%
Vapour quality  0.05
Heat transfer coecient  5.0%
The absolute pressure is also measured at the inlet of
the precondenser, where superheated vapor ¯ows, while
pressure drop across the test section is measured by a
di€erential transducer with an accuracy around 0.06
kPa. The heat ¯ux exchanged is derived from an
enthalpy balance on the cooling water side. Water ¯ow
rates in the test sections and the other tube-in-tube
exchangers are measured by magnetic-type ¯ow meters
with an accuracy around  1% of the measured value.
A Keithley data acquisition system together with a per-
sonal computer are used to acquire, analyze and store
the data.
It was estimated from a propagation of error analysis
that the heat transfer coecients were measured to an
accuracy of  5.0% at typical test conditions. A list of
accuracies for sensors and parameters is reported in
Table 1.
5. Heat transfer results
Condensation tests are carried out in a plain 8.0 mm
ID tube (longitudinal relative roughness was measured
to be equal to 0.0001) at saturation temperature ranging
from 30 to 60 C and mass velocity ranging from 100 to
750 kg/(m2 s). The inlet vapour quality varies between
0.2 and 0.8, and approximately 15±25% vapour quality
change occurs in the test tube depending on the mass
¯ux velocity and heat ¯ux. The heat transfer values
presented are actually mean values over a small change
of vapour quality and can be referred to as quasi-local
values. Only for tests with mass velocity below or equal
to 200 kg/(m2s) the vapour quality change can go up to
35±40% under high driving temperature di€erence.
Data acquisitions of the test variables show that no
signi®cant ¯uctuations or cyclic changes in tempera-
tures, pressures or ¯ow rates occur during tests.
Repeatability of experimental data is very good (within
2%) for tests done on di€erent days. Subroutines cal-
culate test variables in order to run the experiments at
speci®c conditions. A summary of test conditions is
reported in Table 2. In order to analyse the condensa-
tion mechanisms, the following discussion will focus on
Temperature di€erence  0.03 C
Absolute pressure  0.1% f.s.
Di€erential pressure  0.06 kPa
Heat ¯ow rate  4.5%
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87 79

Table 2
Main operating conditions during condensation tests
Tableau 2
Conditions de fonctionnement lors des essais sur la condensation

Fluid Tsat psat,av Ga xm
x
T

( C) (kPa) [kg/(m2s)] (K)

R22 40+1±3 1550 100 0.21±0.64 0.25±0.4 5.5±8

+3±2 1550 200 0.21±0.83 0.2±0.36 5±12
+2±3 1480 400 0.26±0.78 0.18±0.39 5±15
+2±0 1580 600 0.28±0.84 0.2 7±11
+2±3 1490 750 0.23±0.80 0.18±0.28 7±15
R134a 300.5 770 400 0.35±0.82 0.23 5±9
40+3±0 1080 65 0.23±0.71 0.23±0.29 3
+3.5±0 1090 100 0.25±0.79 0.15±0.37 2.5±6
+3±0 1090 200 0.21±0.78 0.23±0.4 6.5±11
+1±0 1040 300 0.25±0.74 0.19±0.31 7±11
0.5 1020 400 0.24±0.77 0.22 6±10
+0±4 920 750 0.29±0.80 0.21 7±13
50+1.5±1 1360 400 0.17±0.75 0.2±0.27 9
60+0±1.5 1680 400 0.30±0.64 0.2±0.26 9
R236ea 30+2±0 260 200 0.24±0.77 0.22 4±6.7
0.5 246 400 0.28±0.78 0.18 3±6.5
401 345 100 0.22±0.75 0.2±0.4 3±6
0.5 340 200 0.18±0.77 0.2±0.23 4.4±6.8
+0.5±2 325 400 0.25±0.85 0.17±0.24 4±7.5
1 340 600 0.23±0.78 0.16±0.23 3.4±8
511 482 200 0.26±0.76 0.2±0.31 5±8
1 480 400 0.15±0.80 0.17±0.2 3.6±9
R125 30+2±0 1600 200 0.23±0.77 0.27 6±7
1 1600 400 0.25±0.77 0.23±0.26 7±9
+0±2 1520 750 0.40±0.79 0.18 6±8
40+0±2 1930 100 0.25±0.67 0.22±0.28 3.5
+0±3 1930 150 0.25±0.59 0.2±0.4 4±9
+1.5±0.5 2040 200 0.27±0.78 0.24±0.28 5.5
1 2050 400 0.24±0.79 0.27 8±10
+0±2 1960 750 0.25±0.79 0.25 8±12
521 2680 200 0.30±0.65 0.25±0.36 6
+3±0 2790 400 0.32±0.82 0.24±0.34 8
1 2670 750 0.34±0.75 0.23 6±8
R32 301 1960 200 0.25±0.65 0.2 5±7
0.5 1930 400 0.30±0.80 0.14±0.22 4±12
0.5 1930 600 0.33±0.81 0.12±0.24 5±14
400.5 2500 100 0.37±0.70 0.2±0.39 3.6±7
0.5 2490 200 0.25±0.82 0.18±0.2 5±7
0.5 2480 400 0.25±0.86 0.15±0.2 5±12
0.5 2490 600 0.25±0.80 0.15±0.2 5±13
500.5 3140 200 0.25±0.83 0.18±0.2 5±7
0.5 3130 400 0.22±0.86 0.15±0.2 5±12
0.5 3150 600 0.28±0.79 0.15±0.24 6±14
R410A 281 1770 400 0.40±0.86 0.15 5±7
40+0.5±1 2380 100 0.37±0.77 0.27±0.4 4.6±7.4
1.5 2420 200 0.21±0.73 0.22±0.39 7±11
+0±2.5 2310 400 0.15±0.88 0.18±0.24 5±12
+0.5±1.5 2390 750 0.25±0.76 0.21±0.25 10±13
501 3070 400 0.22±0.83 0.19±0.26 8
a
Mass velocity varies within 2% for G4200 kg/(m2 s), 1% for G=300400 kg/(m2 s), 0.5% for G=600750 kg/(m2 s). A
wider range applies to R22.
80 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
the main parameters which may have an e€ect on heat
transfer performance: mass velocity, vapour quality,
saturation temperature, driving temperature di€erence,
reduced pressure. The relative importance of these
parameters depends on the test conditions.
Experimental heat transfer coecients for the pure
HFC refrigerants R236ea, R134a, R125, R32 and the
nearly azeotropic mixture R410A are plotted over
vapour quality as a function of mass velocity in Fig. 3
(mass velocity is referred to as G in the ®gure). Only
data referring to a condensation temperature around
40 C are reported in the graphs. R22 experimental data
at the same condensation temperature are also reported
in the same ®gure as the reference refrigerant, and one
can see that some trends are common to all the refrig-
erants shown.
Condensation heat transfer coecient increases with
increasing mass velocity and vapour quality. At a con-
stant value of saturation temperature and mass velocity,
the experimental trend is pretty much linear against
vapour quality at high mass ¯ow rates. For R22, a larger
data scattering is due to the wider range for the mass
velocity and the saturation temperature during tests.
Due to the limited volume of the experimental test rig,
when varying the vapour quality it was not always pos-
sible to reach exactly the same operating pressure for all
the data points.
When mass velocity is higher or equal to 400 kg/(m2
s) no dependence of the heat transfer coecient on the
driving temperature di€erence is found, and all points
taken at a temperature di€erence from 5 to 15 K lay on
the same curve. At low mass velocity [below or equal to
100 kg/(m2s)], there is a less evident dependence of heat
transfer coecient on vapour quality. This trend sug-
gests that the vapour shear forces in these situations are
almost negligible and the ¯ow pattern is strati®ed. At
this mass velocity, on the contrary, an e€ect of the tem-
perature di€erence between saturation and wall can be
observed: the heat transfer coecient increases as
T
decreases. As the mass velocity increases, at ®rst the
¯ow pattern remains basically strati®ed (as indicated by
the ¯ow pattern maps considered further on) but the
vapour shear forces become more and more signi®cant.
This results in a higher rate of increase of heat transfer
coecient with vapour quality, thus approaching the
typical trend of fully shear dominated condensation.
At higher ¯ow rates (400 kg/(m2 s) and over), as
already pointed out, the condensation coecient shows
a linear trend and a higher slope against vapour quality
and no dependence of the heat transfer coecient on

T is found for those test conditions. It can be con-
cluded that a fully developed annular ¯ow is then
occurring and forced convection is the sole driving heat
transfer mechanism.
Condensation heat transfer coecients were also
measured at saturation temperatures varying between
30 and 50 C for R236ea, R125, R32 and R410A, and
from 30 up to 60 C for R134a. The condensation heat
transfer coecient at constant mass velocity can be
approximated by a linear trend against vapour quality,
provided that the mass velocity be high so as to have an
annular ¯ow pattern in the tube. Higher values of heat
transfer coecient are obtained at lower saturation
temperature, mass velocity and vapour quality being the
same.
Useful information about ¯ow patterns can be
obtained by plotting the experimental data on a proper
¯ow pattern map. Breber et al. [4] proposed a ¯ow pat-
tern map, based on the dimensionless vapour mass
velocity JG plotted over the Martinelli parameter Xtt,
where:
JG ˆ x G=‰g d G …L ÿ G †Š0:5 …3†
and x is vapour ¯ow quality, d is tube diameter, L and
G are liquid and vapour density respectively. Breber et
al. divided the map into four main regions, and transi-
tion zones between them. Their map together with the
following one by Tandon et al. [16] have been developed
speci®cally for condensation two-phase ¯ow.
Fig. 5 shows the experimental data plotted on the
Breber et al. [15] map. When the ¯ow pattern is strati-
®ed, a laminar ¯ow of condensate runs down the upper
surface of the tube and collects as a strati®ed layer of
liquid in the bottom part of the tube: it is the case of low
mass velocity ¯ow conditions. When mass velocity
increases, the experimental points move towards the
annular and mist annular zone, implying that forced
convection becomes the sole driving heat transfer
mechanism in this case.
The Breber et al. [15] map predicts annular and mist
annular ¯ow for all data points at mass velocity G equal
to or greater than 400 kg/(m2 s). At lower mass velocity
most of the points lay in the transition region and at the
lowest mass velocity tested [G=65±100 kg/(m2 s)]
almost all points lay in the wavy and strati®ed zone. The
same results are found by adopting the Tandon et al.
[16] map, with JG plotted over a function of the void
fraction e, calculated from the Smith equation [17] (Fig. 6).
When plotting the present data taken at di€erent
saturation temperatures but same mass velocity and
vapour quality (and same ¯uid), no large scattering is
found, which means that, while saturation temperature
strongly a€ects the heat transfer coecient, it does not
show a similar e€ect on the ¯ow pattern.
Experimental tests were conducted at reduced pressure
ranging from 0.07 (as is the case for R236ea condensing at
30 C) to 0.7 (for R125 condensing at 50 C). By looking
at the ¯ow pattern maps, one can see that high pressure
¯uids (with higher reduced pressure values in the
experiments) show higher values of the Martinelli para-
meter (in the Breber map) and of the function (1-")/" in
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87 81

Fig. 3. Heat transfer coecient vs. vapour quality when varying mass velocity for R22, R134a, R236ea, R125, R32, R410A (G is mass
velocity in [kg/(m2s)]). All data reported in Table 2 at approximately Tsat=40 C are plotted in the above graphs: at low mass velocity
the value of
T a€ects the heat transfer coecient.
Fig. 3. Coecient de transfert de chaleur en fonction de la qualite de la vapeur pour divers deÂbits massiques de R22, R134a, R236a, R125,
R32, R410A (G eÂtant le deÂbit massique exprime en [kg/(m2.s]). Toutes les donneÂes representeÂes dans le Tableau 2 aÁ une tempeÂrature
d'environ Tsat = 40 C sont traceÂes dans les courbes ci-dessus : pour les faibles deÂbits massiques, la valeur
T exerce une in¯uence sur le
coecient de transfert de chaleur.
82 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87

Fig. 4. Pressure gradient vs. vapour quality when varying mass velocity for R-22, R-134a, R-236ea, R-125, R-32, R-410A at
approximately Tsat=40 C (G is mass velocity in [kg/(m2 s)]).
Fig. 4. Gradient de pression en fonction du deÂbit massique pour le R22, le R134a, le R236ea, le R125, le R32, le R410A aÁ une tempeÂrature
d'environ Tsat = 40 C (G eÂtant le deÂbit massique exprime en [kg/(m2.s)].
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87 83

Fig. 5. Condensation ¯ow regimes on the Breber et al. [15] map.

Fig. 5. Carte de Breber et al. [15] montrant les eÂcoulements lors de la condensation.

Fig. 6. Condensation ¯ow regimes on the Tandon et al. [16] map.

Fig. 6. Carte de Tandon et al. [16] montrant les eÂcoulements lors de la condensation.
84 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87

the Tandon map. In other words, ¯ow patterns for high
pressure ¯uids move to the right of the maps, towards
the region of bubble ¯ow, which cannot be achieved in
condensation unless the reduced pressure is almost 1
(see Breber [15]).
From Fig. 3, heat transfer performance of the three
families of ¯uids (low-, mid- and high-pressure ones)
can easily be compared at the same mass velocity with
one another. On average, low and mid-pressure refrig-
erants perform better than high pressure refrigerants.
Regarding the heat transfer performance of R410A, it is
higher than that of its component R125, enhanced by
the good thermal properties (liquid phase thermal con-
ductivity, in particular) of R32.
6. Pressure drop results
As explained above, pressure drop during condensa-
tion is measured by using a di€erential transducer, with
an accuracy of  0.06 kPa. In order to avoid any pre-
sence of liquid and to obtain precise measurements, the
connecting lines are heated by electric resistances. It was
observed that heating the transducers above 50 C could
a€ect the reading. The error caused by heating the
transducer was found to be negligible at high mass
velocity, but it can be signi®cant at low measured values
(<500 Pa).
The measured pressure drop regards a condensing
¯uid, so it can be thought as the sum of a frictional
component and a component due to momentum
change. In Fig. 4 pressure drop data are plotted against
the average vapour quality for the same ¯uids and ¯ow
conditions as in the heat transfer plots of Fig. 3. Among
operating conditions reported in Table 2, only data
referring to a condensation temperature around 40 C
are reported in the graphs. The mass velocity is again
referred to as G [kg(m2 s)]. Although the pressure gra-
dient is reported for values down to zero or even
below zero in the ®gure, the instrument accuracy in
these conditions can be comparable with pressure drop
measurements.
Pressure drop increases with increasing vapour qual-
ity and mass ¯ow rate, and higher values are measured
at lower saturation temperature. On average, at the
same mass velocity and vapour quality, low pressure
¯uids show a higher pressure drop penalty. Thus, when
comparing the heat transfer performance of two ¯uids
condensing at the same temperature and mass velocity
but far di€erent operating pressures, one should
account for the di€erent pressure drop behaviour: the
comparison among ¯uids at the same mass velocity is

Fig. 7. Calculated vs. experimental Nusselt number: predictions by Kosky and Staub model [1] for annular ¯ow and Jaster and Kosky
equation [12] for strati®ed ¯ow.
Fig. 7. Nombre de Nusselt calcule en fonction du nombre de Nusselt expeÂrimental : preÂvisions selon le modeÁle de Kosky et Staub [1] pour
l'eÂcoulement annulaire et l'eÂquation de Jaster et Kosky [12] pour l'eÂcoulement strati®eÂ.
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
misleading if it does not account for di€erent saturation
temperature drops as well (see Cavallini et al. [14]).
Experimental data have been compared against the
well known correlation by Friedel [18] for computing
the two-phase frictional multipliers, with reference to
pressure drop inside smooth tubes. In the present work,
the single-phase friction factor used in the above models
is estimated from the Blausius equation. As Friedel
equation has been developed for adiabatic two phase
¯ow, it is here applied by using the correction of Mickley
et al. [19], which increases the predicted frictional pressure
gradient in condensation. The computation of the
momentum pressure change, which involves a pressure
recovery during condensation, requires the evaluation of
the vapour void fraction e: the model by Rohuani [20]
has been used along with that by Friedel, as suggested
by Friedel himself. Results of the comparison are shown
in Fig. 7: the Friedel correlation seems to slightly over-
estimate the experimental pressure drop for high pres-
sure ¯uids and underestimate it for low pressure ¯uids
at high mass velocities.
7. Comparison of heat transfer coecient data against
predicting models
For any prediction of two-phase heat transfer the
occurring ¯ow regime must be considered. As it can be
seen in the ¯ow pattern maps of Figs. 6 and 7, experi-
mental ¯ow conditions presented here cover both annular
and wavy-strati®ed ¯ows. For the case of annular ¯ow,
the experimental heat transfer is compared against the
Kosky and Staub [1] model, which is based on the
momentum-heat transfer analogy. This analysis allows to
calculate the heat transfer coecient during condensation
inside a smooth tube assuming annular ®lm and constant
uniform ®lm thickness along the tube circumference.
Applying the momentum-heat transfer analogy to calcu-
late the liquid layer thermal resistance, one gets the fol-
lowing expression for the heat transfer coecient:
q=A cL L u cL …L  †0:5
ˆ ˆ ˆ …4†
Tsat ÿ Tw T‡ T‡
where ut is the friction velocity,  is the shear stress
(assumed constant in the liquid ®lm), T+  is a dimen-
sionless parameter evaluated at the vapour-liquid inter-
face. Assuming the Von Karman universal velocity
distribution inside a pipe to hold for the liquid con-
densate layer, T+ +
 can be derived as a function of  :
T‡ ‡ ‡
 ˆ  PrL for  45
 ÿ ‡

T‡
 ˆ 5 PrL ‡ 5 ln 1 ‡ PrL  =5 ÿ 1 for 5 < d‡ < 30
ÿ ‡
Kosky and Staub model, higher deviations are found for
T ˆ 5 PrL ‡ 5 ln…1 ‡ 5PrL † ‡ 2:5ln  =30 for ‡ 530
‡
…5†
85
where PrL is the Prandtl number of the liquid phase,
and the dimensionless ®lm thickness ‡ ˆ L u =L .
Kosky and Staub related + to the liquid layer Rey-
nolds number ReL ‡ G d …1 ÿ x†=L :
‡ ˆ …ReL =2†0:5 for ReL 41145
…6†
‡
 ˆ 0:0504Re7=8
L for ReL > 1145
The interfacial shear stress is ®rst determined:
 ˆ …ÿdpf =dz†d=4 …7†
where (-dpf /dz) is the frictional pressure gradient along
the channel and d is the tube inside diameter. The fric-
tional pressure gradient is evaluated by using the mea-
sured pressure drop and subtracting the acceleration
component. Considering that the coordinate z is orien-
ted in the ¯ow direction:
d  2
ÿ…dp=dz†a ˆ G2 x =…G "† ‡ …1 ÿ x†2 =‰L …1 ÿ "†Š
dz
…8†
where " is the void fraction calculated by the Rohuani
model [20]. Because present data are actually taken with
a small change in vapour quality in the total test section
length, the momentum term is much smaller than the
frictional term.
When the ¯ow is not fully annular, the above model
cannot be used since the shear stress is not the sole
driving mechanism anymore. The Jaster and Kosky [12]
equation is then used to calculate the heat transfer
coecient under hypothesis of gravity controlled con-
densation, by adopting again the Rouhani [20] correla-
tion to calculate the void fraction.
After application of the Kosky and Staub analysis
and the Jaster and Kosky model, the higher of the two
values is taken as the predicted heat transfer coecient.
In Fig. 8 calculated Nusselt numbers are plotted versus
experimental values.
The ®gure shows a satisfactory agreement between
predicted and experimental values, with a mean percent
deviation, de®ned from the absolute values of deviations
(MDA), around 12%, where:
X
MDA ˆ …1=Np† ‰jhCALC ÿ hEXP j=hEXP Š 100 …9†
with Np being the total number of experimental points.
Higher deviations are observed at low ¯ow rates [65±200
kg/(m2 s)], when condensation is gravity driven and
prediction is given by the Jaster and Kosky equation as
detailed above. Mean deviations for all data sets are
listed in Table 3. With reference to the results of the
R410A and R32, while deviations for all the other
models are within 10%.
86 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87

Fig. 8. Calculated vs. experimental pressure gradient: predictions by Friedel model [18].
Fig. 8. Gradients de pression calculeÂs ou expeÂrimentaux : preÂvisions selon le modeÁle de Friedel [18].

Together with MDA, another deviation is used here allow to conclude much about the general trend of the
which is the algebraic sum of the relative deviations prediction, MDR with its negative or positive sign
divided by the number of experimental points: shows when the method gives a conservative prediction
X or overestimates the data. As one can see, experimental
MDR ˆ …1=Np† ‰jhCALC ÿ hEXP j=hEXP Š 100 …10† coecients are generally overestimated.

As MDA describes the agreement between the calcu-

lated values and the experimental ones, but does not 8. Conclusions

Heat transfer coecients and pressure drops are

Table 3
Mean percentage deviations (MDR and MDA) between calcu-
measured during condensation inside a smooth tube
lated and experimental heat transfer coecients when operating with HFC refrigerants R236ea, R134a,
R22, R125, R32, R410A. The e€ects of vapour quality,
Tableau 3
Ecarts moyens (MDR et MDA), exprimeÂs en pourcentage, entre
mass velocity, saturation temperature, driving tempera-
les coecients de transfert de chaleur calculeÂs et expeÂrimentaux ture di€erence and reduced pressure are discussed in the
paper.
Refrigerant Kosky and Staub Jaster and Kosky The model by Kosky and Staub and the equation of
model (annular ¯ow) model (strati®ed ¯ow) Jaster and Kosky, valid respectively in the annular ¯ow
MDR MDA MDR MDA regime and in the strati®ed ¯ow regime, can fairly pre-
dict condensation heat transfer coecients even with the
R22 2.4 3.3 ÿ10.7 10.7 new high pressure HFC ¯uids, whereas well established
R134a 3.8 9.0 11.4 22.1 semiempirical equations are not applicable.
R236ea 2.7 6.1 5.5 11.4 For prediction applications, when frictional pressure
R125 ÿ5.2 9.0 ÿ5.6 14.7
drop is not known, the method by Friedel to estimate
R32 20.5 20.9 13.3 13.9
the frictional pressure gradient is suggested in conjunc-
R410A 12.3 13.8 8.7 15.0
tion with the Kosky and Staub procedure.
 A. Cavallini et al. / International Journal of Refrigeration 24 (2001) 73±87
References
[1] Kosky PG, Staub FW. Local condensing heat transfer
coecients in the annular ¯ow regime. AIChE J 1971;17.
[2] Traviss DP, Rohsenow WM, Baron AB. Forced convec-
tion inside tubes: a heat transfer equation for condenser
design. ASHRAE Trans 1972;79:157±65.
[3] Dobson MK, Chato JC. Condensation in smooth hor-
izontal tubes. J Heat Transfer, ASME 1998;120:193±213.
[4] NIST, Refprop, Version 6.01, US Department of Com-
merce, 1999.
[5] Tang L. Empirical study of new refrigerant ¯ow con-
densation inside horizontal smooth and micro-®n tubes.
University of Maryland at College Park, PhD thesis, 1997;
pp. 251.
[6] Ebisu T, Torikoshi K. Experimental study on evaporation
and condensation heat transfer enhancement for R407C
using herringbone heat transfer tube. ASHRAE Trans
1998; 104(2).
[7] Haraguchi H, Koyama S, Fujii T. Condensation of
refrigerants HCFC 22, HFC 134a and HCFC 123 in a
horizontal smooth tube (2nd report, proposal of empirical
expressions for local heat transfer coecient). Trans
JSME 1994;60(n.574):245±52.
[8] Eckels SJ, Tesene BA. A comparison of R22, R134a,
R410A, and R407C condensation performance in smooth
and enhanced tubes: part I, heat transfer. ASHRAE Trans
1999, 105.
[9] Cavallini A, Zecchin R. A dimensionless correlation for
heat transfer in forced convection condensation. 6th Int
Heat Transfer Conf, Tokyo 1974;3:309±13.
87
[10] Shah MM. A general correlation for heat transfer during
®lm condensation inside pipes. Int J Heat Mass Transfer
1979;22:547±56.
[11] Nusselt W. Die Ober¯achenkondensation des Wasser-
dampfes. Z Ver Dt Ing 1916;60:569±75541±546.
[12] Jaster H, Kosky PG. Condensation heat transfer in a mixed
¯ow regime. Int J of Heat Mass Transfer 1976;19:95±9.
[13] Cavallini A, Del Col D, Doretti L, Longo GA, Rossetto
L. Condensation of refrigerants inside plain and enhanced
tubes. Proc. 3rd European Thermal Sciences Conf., Hei-
delberg 2000;1:51±60.
[14] Cavallini A, Del Col D, Doretti L, Longo GA, Rossetto
L. Condensation heat transfer of new refrigerants: advan-
tages of high pressure ¯uids. Proc. 2000 Int. Refrig. Conf.
Purdue, 2000;177±84.
[15] Breber G, Palen J, Taborek J. Prediction of horizontal
tube-side condensation of pure components using ¯ow
regime criteria. J of Heat Transfer 1980;102:471±6.
[16] Tandon TN, Varma HK, Gupta CP. A new ¯ow regime
map for condensation inside horizontal tubes. ASME J
Heat Transfer 1982;104:763±8.
[17] Smith SL. Void fraction in two-phase ¯ow: a correlation
based on equal velocity head model. Heat and Fluid Flow
1971;1( n. 1):22±39.
[18] Friedel L. Pressure drop during gas/vapor-liquid ¯ow in
pipes. Int Chem Eng, July 1980;20:352±67.
[19] Mickley HS, Ross RC, Squyers AL, Stewart WE. Heat,
mass and momentum transfer for ¯ow over a ¯at plate
with blowing or suction, Rep. No. NACA-TN-3208, 1954.
[20] Rouhani SZ. Subcooled void fraction. AB Atomenergi
Sweeden, Internal Rept. AE-RTV841, 1969.